Visual saliency algorithm based on overall color contrast ratio and space distribution in image

Abstract

The invention discloses a visual saliency algorithm based on an overall color contrast ratio and space distribution in an image. The visual saliency algorithm based on the overall color contrast ratio and the space distribution in the image comprises the following steps: (1) inputting an original image, pre-dividing the original image into N areas through a mean shift algorithm, wherein i is larger than or equal to 1 and smaller than or equal to N, generating N area mark images after pre-dividing, (2) calculating a color saliency value of an ith area, (3) calculating color saliency values of pixels, (4) calculating a position saliency value of the ith area, (5) calculating position saliency values of the pixels, (6) calculating the color saliency values of the pixels and the position saliency values of the pixels, carrying out normalization on the saliency values of the pixels, and calculating a normalized saliency value of each pixel. According to the visual saliency algorithm based on the overall color contrast ratio and the space distribution in the image, the overall color contrast ratio and the space distribution are combined, a saliency image can be calculated, wherein the resolution of the saliency image is the same as that of the original image resolution, and saliency objects in the calculated saliency image are evenly salient. Meanwhile, a background is well restrained, and the visual saliency algorithm based on the overall color contrast ratio and the space distribution in the image is more suitable for content-based application occasions such as image division.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a method for calculating visual saliency in images. Background technique [0002] The calculation process of visual saliency is to simulate the process of human eyes observing the image, and then extract the area of ​​interest of human eyes, and finally obtain a visual saliency map corresponding to the degree of attention. When the human eye observes an image, it often pays attention to a relatively prominent area or image block, and these prominent areas or image blocks are called visually salient areas. The process of highlighting these visually salient regions by a certain calculation method is called visually salient region detection. In order to facilitate efficient subsequent processing of images, inspired by the attention mechanism of the human eye, we use a computer to simulate the human visual system to detect visually salient regions of the image. In many aspects such a...

AI Technical Summary

Problems solved by technology

But the disadvantage is that the resolution of the obtained visual saliency map is lower than that of the original image, resulting in the loss of original image information.

The saliency detection method proposed by Goferman (Reference: S Goferman, L Zelnik-Manor, A Tal. Context-aware saliency detection. IEEE Conf. on CVPR, 2010) solves the problem of low resolution while taking into account the local differences of the image and the global difference, and deliberately preserve the background area around the salient object, but Goferman's method usually produces higher saliency values ​​near the edge of the object, while the saliency value inside the object decreases, and evenly highlights the entire There are still deficiencies in visually salient objects

Images